摘要
自然资源领域是技术密集型领域,面向新时代、新任务、新要求,迫切需要应用高新技术构建调查监测技术体系,支撑服务自然资源调查、监测、评价、规划等工作。为了向国土空间开发利用和国土空间治理提供高质量的数据供给,立足人工智能技术发展现状,以卫片执法遥感监测、全国地类变化遥感监测等国家重大战略需求为牵引,针对传统变化检测方法人工劳动繁重和经典变化检测方法工序复杂的固有问题,本文提出了一种网格化人机协同生产方法。通过将遥感影像待检测区域进行北斗网格剖分,构建最小监测单元,规范监测作业模式;创新了一种多模态变化检测技术,以集群处理技术为计算中枢,将多种人工智能技术优势进行融合互补;研究了分布式遥感变化推理引擎,突破机器动态变化感知能力,使机器可以快人一步,动态检测操作视图影像变化,实现人与机器的同步变化感知;探索了人机协同信息化监测作业体系,基于北斗网格位置码建立监测任务调度管理看板,使监测任务进度管理精细可控,实现了作业效率30%和作业质量20%的提升。
The field of natural resources is a technology-intensive field.Facing the new era,new tasks,and new requirements,there is an urgent need to apply high-tech to build a survey and monitoring technology system to support and serve natural resource surveys,monitoring,evaluation,and planning.In order to provide high-quality data supply for the development and utilization of land space and land space governance,based on the development status of artificial intelligence technology,guided by national major strategic needs such as satellite law enforcement remote sensing monitoring and national land type change remote sensing monitoring,the traditional change detection method Due to the inherent problems of heavy manual labor and complex procedures of the classic change detection method,a grid-based man-machine collaborative production method is proposed.By dividing the remote sensing image area to be detected into a BeiDou grid,the smallest monitoring unit is constructed to standardize the monitoring operation mode.;Innovated a multi-modal change detection technology,using cluster processing technology as the computing center,integrating and complementing the advantages of various artificial intelligence technologies;researched the distributed remote sensing change reasoning engine,breaking through the machine,s dynamic change perception ability,so that the machine can One step faster,dynamic detection of image changes in the operation view,realizing the synchronous change perception of man and machine;exploring the man-machine collaborative information monitoring operation system,and establishing a monitoring task scheduling management kanban based on the BeiDou grid location code,so that the monitoring task progress management is fine It is controllable,achieving a 30%improvement in operating efficiency and 20%in operating quality.
作者
刘立
董先敏
王德富
张志强
刘娟
LIU Li;DONG Xianmin;WANG Defu;ZHANG Zhiqiang;LIU Juan(The Third Geographical Information Mapping Institute of Natural Resources Ministry,Chengdu 610100,China)
出处
《测绘通报》
CSCD
北大核心
2024年第S01期130-136,共7页
Bulletin of Surveying and Mapping
基金
自然资源部四川省滑坡灾害隐患遥感智能防控体系部省合作研究项目(SCDZRS2023)
关键词
多模态
变化监测
人机协同
深度学习
自然资源调查监测
multi-modal
change monitoring
human-machine collaboration
deep learning
natural resource survey and monitoring
作者简介
刘立(1989—),男,工程师,从事自然资源调查监测与地理信息工程应用研究。E-mail:274114486@qq.com
